PAPERZILLA

High-signal always up-to-date research data feeds

Research feeds for people and agents

AI Agents welcome.▮

Connect an agent to a Paperzilla project. New papers are monitored and ranked against your research question, so every run starts with an up-to-date feed.

Claude · Codex · OpenClaw · Cursor · your agent

Create a research feed Read agent setup docs ↗

Use the same project in the app, by email, through MCP or the CLI, and as an RSS/Atom feed.

Why a research feed

A Paperzilla project keeps your topic, filters, and feed between sessions.

The project stores the topic, categories, include and exclude keywords, plus the matching papers found over time.

Paperzilla handles monitoring and relevance ranking. Your agent can read the feed, compare papers, prepare briefs, and send them where they are needed.

Example Paperzilla project

Project

Agents evaluation

Intent

How to evaluate coding agents

Signal

Must Read + Related

State

Continuously updated

Monitoring loop

How a project stays current.

Paperzilla checks the sources below and updates the ranked project feed as new papers appear.

Set up the project

Add a topic, choose categories, and set any include or exclude keywords. Use a focused project for each research objective.

Paperzilla checks new papers

Each new paper is evaluated against the project criteria and filed as Must Read or Related.

Connect the agent

Use MCP, the CLI, or RSS/Atom to browse new recommendations, search the project history, and retrieve paper context.

Monitored sources

arXiv

bioRxiv

medRxiv

ChemRxiv

ChinaXiv

PubMed

Agent capabilities

What an agent can do with a Paperzilla project.

Through MCP or the CLI, an agent can search the full feed and retrieve paper metadata or Markdown.

Projects

List projects and inspect the topic, mode, visibility, and project ID before the agent starts reading.

Recent feed items

Pull recent recommendations and filter by Must Read status, date, or result limit.

Full-feed search

Server-side search covers titles, authors, abstracts, and summaries across the whole project feed. Feedback filters are available too.

Paper and recommendation records

Retrieve either the canonical paper record or its project recommendation, including relevance, note, and feedback context.

Paper markdown

Retrieve paper markdown when ready or queue its preparation when project context allows it, with a structured status.

Atom URL

Use the private, tokenized project URL with a feed reader, poller, or existing feed infrastructure.

Paperzilla MCP is read-oriented. Use the CLI when an automation should leave Useful, Very useful, or Not for me feedback on a recommendation.

Connection options

Choose how the agent connects.

MCP, the pz CLI, RSS/Atom, and the app all read from the same Paperzilla project.

Native tool calling

MCP

Use MCP for native tool calling in supported assistants. The Paperzilla MCP endpoint is read-oriented.

projects_list · feed_search · paper_markdown

Connect with MCP ↗

Shell + structured output

pz CLI

Use the pz CLI for shell scripts, JSON output, local agents, and project-specific feedback.

pz feed search --project-id … --json

Install the CLI ↗

Passive polling

RSS / Atom

Use RSS/Atom to poll a private, tokenized project feed URL.

one secure URL per user + project

Use a project feed ↗

Human review

App + email

Use the app for triage and email for daily or weekly digests from the same project.

Must Read · Related · Useful · Very useful

See human workflows ↗

Documented setups

Connect Claude, Codex, OpenClaw, or another agent.

Use MCP for native tool calls or the pz CLI for shell-based workflows.

OpenClaw

Skill + CLI

The published Paperzilla skill on ClawHub teaches OpenClaw how to use an installed and authenticated pz CLI. Once configured, OpenClaw can use it in chat, cron jobs, and channel delivery.

Setup guide ↗

Claude

MCP or CLI

Use the Paperzilla MCP connector in Claude, Claude Desktop, or Cowork if your plan supports custom connectors. Claude Code can also run pz directly.

Setup guide ↗

Codex

MCP or CLI

Connect Paperzilla through Integrations & MCP, or let Codex run pz in a shell. Use a Codex automation to prepare a recurring brief.

Setup guide ↗

Your agent

MCP or CLI skill

Cursor and other Streamable HTTP MCP clients can use the endpoint. AgentSkills-compatible clients can use the CLI skill when pz is installed and authenticated.

Setup guide ↗

Channel workflows

Send research briefs to Telegram, Slack, or Teams.

Paperzilla provides the project feed and relevance ranking. OpenClaw, Claude, or Codex can turn the results into a brief and pass it to a channel connector.

Start by reviewing drafts. Direct posting needs a configured connector with permission to post in the channel.

Agent layer

Slack

Microsoft Teams

OpenClaw

Paperzilla skill + pz CLI

Direct when configured

Chat with OpenClaw, schedule a weekday cron job, and have OpenClaw deliver the brief.

Workflow details ↗

Slack

Direct when configured

Use the same CLI and scheduling pattern with an OpenClaw Slack channel.

Workflow details ↗

Microsoft Teams

Setup required

Available after the Teams channel or plugin is configured. The setup has more steps.

Workflow details ↗

Claude + Cowork

Paperzilla MCP

Bridge required

Claude can prepare the brief. Claude Code Channels, OpenClaw, or another bridge sends it to Telegram.

Workflow details ↗

Slack

Documented workflow

Discuss papers in Slack. Cowork can schedule a draft for review before it is sent.

Workflow details ↗

Microsoft Teams

Connector required

Use Microsoft 365 context to prepare a Teams brief. Direct delivery needs a connector.

Workflow details ↗

Codex

Paperzilla MCP or pz CLI

Bridge required

Create a Telegram-ready brief, then send it manually or through your own automation or OpenClaw.

Workflow details ↗

Slack

Setup required

Have Codex prepare the brief. Confirm that a Slack-started task can reach Paperzilla before scheduling it.

Workflow details ↗

Microsoft Teams

Draft workflow

Use a Codex automation to put a Teams draft in Triage. Review it before posting.

Workflow details ↗

Use cases

Common research workflows.

Daily field monitoring

Review the latest Must Read papers in one field without rebuilding a search every day.

Topic watchlists

Keep a separate, tunable project for each major research direction.

Executive or lab briefings

Prepare weekly briefs from a project’s Must Read papers for labs, leaders, or team channels.

Methods and benchmarks

Track methods, model families, datasets, benchmarks, or experimental paradigms.

Competitive and adjacent work

Watch nearby research that could change a roadmap or reveal an unexpected direction.

Agent pipelines and automation

Feed ranked project data into scripts, scheduled tasks, and research agents.

Reading backlog triage

Use relevance labels and structured summaries to choose what to read deeply, skim, or skip.

Try these prompts

Prompts for your Paperzilla project.

Use the project name and explain what the answer will be used for.

Newest papers

“Show the newest Must Read papers from my Agents evaluation project.”

Full-feed search

“Search my project for papers about latent retrieval, including older recommendations in the feed.”

Deep read

“Fetch the markdown for this paper. Summarize its contribution, method, results, and limits, then explain why it matters for our work.”

Compare

“Compare the two most recent Must Read papers from this week. Separate paper metadata from your interpretation.”

Team brief

“Draft a concise weekday research brief for our team channel. Include what you checked, why each selected paper matters, and say clearly if nothing new qualifies.”

For agents reading this

Documentation an agent can read directly.

Every docs page is available as Markdown. The site also publishes LLM indexes and a Streamable HTTP MCP endpoint.

Markdown documentation

Every docs page has a Markdown version that an agent can read directly.

/guides/agent-workflows.md ↗

Documentation indexes

Use the compact docs index for discovery or the full export when the agent needs the whole documentation corpus.

/llms.txt ↗/llms-full.txt ↗

Remote MCP endpoint

Connect a Streamable HTTP MCP client with a personal Paperzilla MCP key. Treat the key like a password.

MCP endpoint docs ↗

FAQ

Common questions.

Is Paperzilla for humans or AI agents?

Both. Humans can use the app feed and email digests while agents and scripts use MCP, the Paperzilla CLI, or RSS/Atom from the same research project.

Which AI agents work with Paperzilla?

Paperzilla documents workflows for Claude, Claude Desktop, Cowork, Claude Code, Codex, OpenClaw, Cursor, and other Streamable HTTP MCP or AgentSkills-compatible clients.

What can an AI agent do with Paperzilla?

An agent can list projects, browse ranked feeds, search a full project feed, inspect paper metadata, retrieve or queue paper markdown, and obtain an Atom feed URL. The CLI also supports project-specific feedback.

Can Paperzilla send research briefs to Telegram, Slack, or Microsoft Teams?

Paperzilla supplies the project feed and relevance ranking. OpenClaw, Claude, or Codex can prepare briefs for those channels. Direct posting depends on the agent, connector, permissions, and channel setup.

Is the Paperzilla MCP connection read-only?

Yes. Paperzilla MCP is read-oriented: it does not edit projects or feedback. It can retrieve project data and may queue paper markdown preparation. Use the CLI when an automation needs to leave project-specific feedback.

Connect an agent to a Paperzilla project.

Create one focused Paperzilla project, then connect it to the agent and channel your team already uses.

Create a research feed Open agent docs ↗

Treat MCP keys and tokenized RSS/Atom URLs like passwords.

Paperzilla

Product

Product Overview AI Agents All Features Pricing Documentation

Resources

Public Digests News Newsletter Paper Summary

Company

About Press Contact

Legal

Terms Privacy